GAF-MN: A New HRRP Target Recognition Method Based on Gramian Angular Field and Matching Networks in Few-Shot Condition

Abstract

Radar high resolution range profiles(HRRP) can provide lots of useful information for radar automatic target recognition(RATR). Recently HRRP based RATR methods adopting deep neural networks have achieved promising performance because of their strong generalization ability. However, these deep models usually need lots of training samples to optimize the parameters, otherwise they would possibly encounter serious overfitting problem in few-shot condition. In this paper, a novel HRRP few-shot recognition method named gramian angular field matching networks(GAF-MN) is proposed to solve the above problems. In the proposed method, a gramian angular field(GAF) technique is adopted to transform the one-dimensional HRRPs to two-dimensional GAF images without any information lost. Further more, the single-channel matching networks is designed to fit the one-dimensional statistical HRRP data. The experimental results demonstrated that the proposed method improved the recognition accuracy and generalization performance in few-shot condition. The proposed method achieved improved performance on an airplane electromagnetic calculation dataset compared with traditional deep learning methods.